Generative AI: Why We Need More Than Common Analogies
Describe The Transformative Character Of Generative AI When Analogies Fall Short
Simplifying Generative AI With Analogies
We are currently witnessing the global roll-out of one of the most transformative technologies of our time: Generative AI. As generative AI rushes through the hype cycle at lightning speed, expert opinions evolve from disregard to excitement, discussion, skepticism, to realism — all within a matter of months, unlike previous generations of AI technology. The genie is out of the bottle and there’s no turning it back inside.
Business and technology leaders are getting excited about the opportunities that generative AI presents. They are looking for ways to adopt the technology quickly with AI leaders often at the receiving end of it. But how can leaders convey the opportunities, risks, and complexities of something as abstract as AI?
Lately, I’ve seen several analogies that seek to compare the impact of generative AI to other transformative innovations of their time. But whichever analogy you pick, it is incomplete and trivializes key characteristics that make generative AI truly unique.
» Watch the latest episodes on YouTube or listen wherever you get your podcasts. «
Why We Need Analogies
Generative AI presents a unique combination of three key characteristics, unlike any other technology before it: Access, Quality/Impact, and Scale. In essence, anyone can use it to create results that are indistinguishable from human-created and examples and distribute them globally in an instant. But because (generative) AI is a rather abstract concept, leaders tend to draw analogies to something the broader public is already familiar with. “Generative AI is like…” — but these analogies rarely capture the full extent. Let’s look at two examples: the printing press and nuclear energy.
Generative AI Is Like…
…The Printing Press (invented ~1440) which has made it possible to create and share information at a broader scale than handwriting ever could.
Access: ✅
Quality/Impact: ✅
Scale: ❌
Unlike the physical output of a printing press, the output of generative AI is an intangible, digital product. The ability to replicate models and digital products and to distribute them globally (and instantly) makes it extremely difficult to control AI — whether it is the LLMs (as equivalents of the printing press) or the applications which embed them (as equivalents of printed papers and books).
To own and operate a printing press, you need to invest money and time to acquire skills, while distribution and scale are still rather limited due to output of a physical product (e.g. paper or book). However, to use generative AI, you don’t need to have a PhD or be a highly trained data scientist. Whether it is knowing how to call an API or even to write plain English, access to the technology is a lot more broadly available.
…Nuclear Energy (invented in the 1940s) which has created a “cleaner” and more efficient source of energy compared to coal.
Access: ❌
Quality/Impact: ✅
Scale: ❌
Nuclear energy does have tremendous potential, while also holding significant dangers if the ingredients are not properly controlled and handled. Any nuclear substance is a physical good whose distribution is inherently limited by the constraints of physics. Plus, globally, access to radioactive raw material is tightly restricted. For example, unlike AI, you cannot buy Uranium — even have access to it for free — and experiment with it in your basement.
Generative AI is an intangible, digital good. Already today, consumer-grade applications using AI can be easily created — without requiring advanced degrees and years of training. Unlike nuclear energy and material, you can also sell these AI-enabled applications and do so to nearly anyone in the world in an instant. And on top of it, Open Source LLMs are publicly available, making it hard to regulate it after the fact.
How Extending Analogies Can Advance The Discourse
Analogies help us to orient ourselves based on similar historic events. At the same time, they can only partially capture the extent of new technologies and their impact. Generative AI is the latest example falling into this pattern as historic, transformational technologies do not exhibit the same characteristics — namely Access, Quality/Impact, and Scale.
We can borrow from historic events, though: Yes, generative AI does have the potential to make information even more rapidly available (like the printing press). And yes, generative AI can lead to adverse outcomes if it is used in the wrong hands (like nuclear energy). But all of the examples will most likely not happen in the same way as previous technologies.
Instead of completely accepting an analogy, leaders can rather describe which aspects of a new technology like generative AI it is similar to. Additionally, leaders can extend the analogy with those aspects which make generative AI unique.
Generative AI…
“…has the potential to transform the world of publishing and information access like the printing press. However, it will enable this more rapidly via the internet.”
“…exhibits opportunities and risks for humankind similar to nuclear energy and substances. However, foundational technologies and models are already publicly accessible.”
Additional analogies that you might have come across include the Internet and social media.
Which generative AI analogy have you come across lately?
» Watch the latest episodes on YouTube or listen wherever you get your podcasts. «
What’s next?
Appearances
June 8 - Panel discussion with Transatlantic AI eXchange on Web 3.0 Generative and Synthetic Data Application
Join us for the upcoming episodes of “What’s the BUZZ?”:
June 8 - Ravit Dotan, Director The Collaborative AI Responsibility Lab at University of Pittsburgh, will join when we cover how responsible AI practices evolve in times of generative AI.
June 20 - Aurélie Pols, Data Privacy Expert & Advisor, will join as we discuss how leaders can shape their business’ accountability for generative AI.
July 6 - Abi Aryan, Machine Learning Engineer & LLMOps Expert, will share how you can fine-tune and operate large language models in practice.
Follow me on LinkedIn for daily posts about how you can set up & scale your AI program in the enterprise. Activate notifications (🔔) and never miss an update.
Together, let’s turn hype into outcome. 👍🏻
—Andreas